Optimization of CO₂ absorption rate for environmental applications and effective carbon capture

Reliable predictions of CO2 absorption are essential for designing effective carbon capture and sequestration (CCS) systems, which are paramount in combating climate change and its health impacts. Artificial Intelligence (AI) and CCS modelling significantly enhance the efficiency and accuracy of CCS technologies, optimizing operations and reducing greenhouse gas emissions. This study presents several standalone machine learning (ML) models for predicting CO2 absorption rates, a critical component in CCS technologies. Subsequently, ensemble models such as Gaussian Process Regression (GPR-SA), Support Vector Machine (SVM-SA), and Random Forest (RF-SA) were evaluated against single models during both training and testing phases. The predicted outcomes were evaluated using various statistical performance indices and 2D-comparative visualization. The study also employed feature engineering to understand the dominancy of the input variables. The prediction skills justified GPR-M4 (94%) emerged as satisfactory and outperformed other model combinations (RF and SVM). The predictive skill during the testing phase, the ensemble models’ robustness, was further featured, with GPR-SA achieving 91% goodness-of-fit despite a slight increase in MAE = 0.0392 and PBIAS = −0.0527. Further quantitative analysis indicated that SVM-SA and RF-SA showed improvements in the testing phase, indicating a strong generalization capability, with SVM-SA reporting an MAE = 0.0065 and a PBIAS = 0.0213 and RF-SA attaining an impressive MAE = 0.0013 and a PBIAS = −0.0703, proposed a slight tendency to underpredict the target variable. Rigorous reliability testing was conducted using the Augmented Dickey-Fuller (ADF), Phillips-Perron (PP), and Jarque-Bera tests to ensure data integrity. The ADF and PP tests confirmed the stationarity of variables, while the Jarque-Bera test indicated that the data followed a normal distribution across all parameters. These findings have significant environmental and public health implications. The study emphasizes the potential of ensemble models in enhancing the accuracy and reliability of CO2 absorption predictions, emphasizing their importance in informing policy decisions and operational strategies for sustainable environmental management.